Nucleoside Salvage Pathway Kinases Regulate Hematopoiesis by Linking Nucleotide Metabolism with Replication Stress

Article

Nucleoside salvage pathway kinases regulate hematopoiesis by linking nucleotide metabolism with replication stress

Wayne R. Austin,1,2 Amanda L. Armijo,1,2 Dean O. Campbell,1,2 Arun S. Singh,1,2 Terry Hsieh,1,2 David Nathanson,1,2 Harvey R. Herschman,1,3 Michael E. Phelps,1 Owen N. Witte,1,4 Johannes Czernin,1,2 and Caius G. Radu1,2

1Department of Molecular and Medical Pharmacology, 2Ahmanson Translational Imaging Division, 3Department of Biological Chemistry, University of California, Los Angeles, Los Angeles, CA 4Eli and Edythe Broad Center of Regenerative Medicine and Stem Cell Research, University of California, San Francisco, San Francisco, CA

Nucleotide deficiency causes replication stress (RS) and DNA damage in dividing cells. How nucleotide metabolism is regulated in vivo to prevent these deleterious effects remains unknown. In this study, we investigate a functional link between nucleotide deficiency, RS, and the nucleoside salvage pathway (NSP) enzymes deoxycytidine kinase (dCK) and thymidine kinase (TK1). We show that inactivation of dCK in mice depletes deoxycytidine triphosphate (dCTP) pools and induces RS, early S-phase arrest, and DNA damage in erythroid, B lymphoid, and T lymphoid lineages. TK1/ erythroid and B lymphoid lineages also experience nucleotide deficiency but, unlike theirdCK / counterparts, they still sustain DNA replication. Intriguingly, dCTP pool depletion, RS, and hematopoietic defects induced by dCK inactivation are almost completely reversed in a newly generated dCK/TK1 double-knockout (DKO) mouse model. Using NSP-deficient DKO hematopoietic cells, we identify a previously unrecognized biological activity of endogenous thymidine as a strong inducer of RS in vivo through TK1-mediated dCTP pool depletion. We propose a model that explains how TK1 and dCK “tune” dCTP pools to both trigger and resolve RS in vivo. This new model may be exploited therapeutically to induce synthetic sickness/lethality in hematological malignancies, and possibly in other cancers.

CORRESPONDENCE Replication stress (RS), a common source generate ribonucleotides that are required for

The Journal of Experimental Medicine Caius G. Radu: of DNA damage and genomic instability RNA synthesis, energy storage, and signal [email protected] (Halazonetis et al., 2008), can be caused by transduction (Evans and Guy, 2004). A fraction Abbreviations used: ATR, deoxyribonucleotide triphosphate (dNTP) de- of the cellular pool of ribonucleotides is con- ataxia telangiectasia and Rad3- ficiency. For example, pharmacological modu- verted into deoxyribonucleotides by ribonu- related protein; dCK, deoxy- lators of dNTP synthesis such as hydroxyurea cleotide reductase (RNR; Fairman et al., 2011). cytidine kinase; DDR, DNA damage response; DKO, double and 5-fluorouracil induce RS (Gagou et al., Nucleoside transporters and deoxyribonucleo- KO; LC-MS/MS, liquid chro- 2010; Arlt et al., 2011). RS is also triggered by side (dN) kinases enable recycling of extracel- matography-tandem mass spec- overexpression of ras and cyclin E, which pro- lular dNs originating from DNA degradation trometry; NSP, nucleoside salvage pathway; RS, replication mote cell division in the absence of sufficient in apoptotic cells (Arnér and Eriksson, 1995), stress; RNR, ribonucleotide dNTP pools to complete genome replication liver biosynthetic processes (Fustin et al., 2012), reductase; RSR, RS response; (Bester et al., 2011). How nucleotide metabo- and food intake. The metabolic flux through SSL, synthetic sickness/lethality; lism is regulated in rapidly dividing cells to the NSP is regulated by rate-limiting kinases. TK1, thymidine kinase 1; UDP, uridine diphosphate. maintain balanced dNTP pools and to prevent Deoxycytidine kinase (dCK) phosphorylates RS is not well understood. Mammalian cells synthesize dNTPs either © 2012 Austin et al. This article is distributed under the terms of an Attribution– Noncommercial–Share Alike–No Mirror Sites license for the first six months de novo or via the nucleoside salvage pathway after the publication date (see http://www.rupress.org/terms). After six months (NSP; Fig. 1 A; Reichard, 1988). The de novo it is available under a Creative Commons License (Attribution–Noncommercial– Share Alike 3.0 Unported license, as described at http://creativecommons.org/ pathway utilizes glucose and amino acids to licenses/by-nc-sa/3.0/).

The Rockefeller University Press $30.00 J. Exp. Med. 2012 Vol. 209 No. 12 2215-2228 2215 www.jem.org/cgi/doi/10.1084/jem.20121061 Figure 1. dCK inactivation causes severe RS in lymphoid and erythroid lineages. (A) Schematic of the de novo pathway (blue) and of the nucleoside salvage pathway (NSP, red) inputs into pyrimidine dNTP pools for DNA synthesis. Solid arrows indicate single-step processes; dashed arrows indicate multistep processes with intermediates not named/depicted in the schematic. Gln, glutamine; Asp, aspartate; UDP, uridine diphosphate; CDP, cytidine disphosphate; TYMS, thymidylate synthase; DCTD, dCMP deaminase; dCK, deoxycytidine kinase; TK1, thymidine kinase 1. (B) dCTP and dTTP pools in WT (black bars) and dCK/ (gray bars) cells; DN Thy, CD4/CD8 double-negative thymocytes; B cells, BM-resident B cell progenitors; Eryth, BM-resident erythroid progenitors. Data are mean values ±SEM for three independent measurements generated from four pooled mice per genotype during each independent measurement. *, P < 0.003; **, P < 0.02. (C) Western blot detection of pChk1 (phosphorylated on Ser345) in lysates from lymphoid and eryth roid progenitors. Total CHK1 protein was used as a loading control. (D) Representative examples of total DNA content staining and percentage of cells / in the G1, S, and G2/M phases in WT and dCK DN3b thymocytes, Hardy fraction B-C cells, and EryA cells. (E) Representative BrdU incorporation into WT

2216 Regulation of replication stress by dCK and TK1 | Austin et al. Article

deoxycytidine to produce deoxycytidine monophosphate system for T lymphocyte development (Taghon et al., 2005). (dCMP), a precursor of both dCTP and dTTP pools (Sabini Collectively, our in vivo and in vitro studies demonstrate an et al., 2008). Deoxyadenosine and deoxyguanosine can also antagonistic functional relationship between dCK and TK1 be phosphorylated by dCK, albeit with significantly lower in regulating dCTP pools in vivo and reveal a previously efficacy than deoxycytidine (Sabini et al., 2008). Thymidine unappreciated role for endogenous thymidine as a biologi- kinase 1 (TK1) phosphorylates thymidine to generate dTMP, cally active metabolite that is linked to the induction of RS a precursor of thymidine triphosphate (dTTP) pools (Arnér in major hematopoietic lineages. and Eriksson, 1995). Although de novo dNTP synthesis is critically important RESULTS for normal DNA replication and repair (Niida et al., 2010; dCTP pool depletion, RS induction, early S-phase arrest, D’Angiolella et al., 2012; Poli et al., 2012; Pontarin et al., and DNA damage in dCK/ lymphoid and erythroid 2012), the role played by the NSP in maintaining the prolifer- hematopoietic lineages ating capacity and genomic integrity of dividing cells is not To investigate the possibility that nucleotide deficiency causes well understood. In vitro studies using transformed cells and the hematopoietic abnormalities identified in the dCK/ exogenous genotoxic agents demonstrated that both dCK and mice (Toy et al., 2010), we quantified dNTP pools in CD4/ TK1 are involved in RS and DNA damage responses (DDR; CD8 double-negative (DN) thymocytes, BM-resident B cell Matsuoka et al., 2007; Chen et al., 2010). However, the pre- progenitors, and in nucleated erythroid progenitors in WT and cise mechanisms connecting dCK and TK1 to RS and DDR dCK/ mice. dCTP pools were significantly reduced in all pathways are yet to be characterized. It is also unknown three dCK/ hematopoietic lineages (Fig. 1 B; *, P < 0.003; whether dCK and TK1 function in vivo to prevent endoge- **, P < 0.02); dTTP, dGTP, and dATP pools were largely nous RS and DNA damage induction in proliferating primary unaffected (Fig. 1 B and Table 1). We next asked whether cells. Analyses of dCK/ (Toy et al., 2010) and TK1/ dCTP deprivation triggers RS in dCK/ hematopoietic (Dobrovolsky et al., 2003) mice support the existence of a lineages by probing for the presence of the activated form functional link between the NSP, RS, and hematopoietic de- of the key ataxia telangiectasia and Rad3-related protein velopment. Studies from our group (Toy et al., 2010), and (ATR) effector and the RS response (RSR) regulator CHK1 confirmed independently by Choi et al. (2012), document kinase (Branzei and Foiani, 2008). Compared with WT cells, severe developmental abnormalities affecting dCK/ T cell, dCK/ lineages displayed significantly higher levels of CHK1 B cell, and erythroid lineages. TK1/ mice also display hema- phosphorylated on Ser345 (pChk1), indicating exposure to topoietic defects characterized by slightly abnormal secondary endogenous RS in vivo (Fig. 1 C). lymphoid structures and by elevated levels of micronucleated To examine the functional consequences of dCTP defi- erythrocytes (Dobrovolsky et al., 2003, 2005). ciency and RS, we analyzed the cell cycle profiles of highly In this work, we sought to elucidate the biochemical and proliferative subpopulations that we found to be selectively molecular mechanisms responsible for the hematopoietic depleted (e.g., T and B cell precursors) or overrepresented phenotypes induced by dCK and TK1 inactivation. We mea- (erythroblast precursors) among dCK/ hematopoietic sured dNTP pools, cell cycle kinetics, and RS levels in lym- progenitor populations (Fig. S1). These highly proliferative phoid and erythroid lineages from mice lacking dCK or TK1. subpopulations included DN3b T cell precursors (CD4, dNTP deficiency and RS induction were detected in both CD8, CD44, CD25med-lo, and CD27hi thymocytes; Taghon dCK and TK1 KO models, prompting us to generate dCK/ et al., 2006), Hardy fraction B-C B cell progenitors (IgM, TK1 double KO (DKO) mice to analyze RS dynamics in B220+, CD43hi, and CD19hi; Hardy et al., 2007), and highly proliferating NSP-deficient hematopoietic precursors. nucleated erythroblast cells (EryA; Ter119+, CD71+, and To our surprise, concomitant inactivation of both dCK and FSChi; Liu et al., 2006). dCK/ cells from all three pro- TK1 ameliorated, rather than aggravated, the dCK/ hema- genitor populations showed abnormal cell cycle profiles topoietic phenotype. To gain additional mechanistic insight (Fig. 1 D), with dCK/ EryA cells displaying the most into how TK1 inactivation rescued the dCK/ RS pheno- pronounced increase in the percentage of cells in S-phase. type, we analyzed dCK/, TK1/, and dCK/;TK1/ Each of the three subpopulations appeared to be arrested in DKO T cell precursors in the OP9-DL1 co-culture model early S-phase (Fig. 1 D).

/ and dCK DN3b thymocytes 1 h after injection of BrdU. Percent of cells in G1, S, and G2 phases of the cell cycle are indicated. (F) Anti-BrdU FITC mean / fluorescent intensities of S-phase cells (SFITC-MFI) from WT and dCK DN3b thymocytes, Hardy fraction B-C cells, and EryA cells. Data are mean values ±SEM. n = 4 mice/genotype; *, P < 0.01. (G) Representative BrdU detection in WT and dCK/ DN3b thymocytes after 3 and 5 h of BrdU chase (n = 4 mice/ genotype). Percentages of cells present in chase gap and BrdU S-phase gates are indicated. (H) Detection of H2A.X phosphorylated on Ser139 (pH2A.X) in DN3b thymocytes by flow cytometry. Percentages of pH2A.X-positive (pH2A.X+) cells are indicated. (I) Quantification of pH2A.X-positive staining in DN3b thymocytes, Hardy B-C cells, and EryA cells from WT (black bars) and dCK/ (gray bars) mice. n = 7 mice/genotype; *, P < 0.0001. (J) S-phase durations, in hours, calculated using 5-h BrdU chase conditions in DN3b thymocytes, Hardy B-C cells, EryA cells, and BM-resident myeloid cells (CD11b+) from WT and dCK/ mice. Data are mean values in hours ±SEM. n = 4 mice/genotype.

JEM Vol. 209, No. 12 2217 Table 1. dATP and dGTP pools (pmol/106 cells) in WT and dCK/ cells dATP pmol/106 cells (±SEM) dGTP pmol/106 cells (±SEM) DN Thy B cells Erythroid DN Thy B cells Erythroid WT 1.08 ( 0.62) 0.64 (0.37) 3.71 (2.15) 3.93 (2.27) 0.76 (0.44) 1.47 (0.85) dCK/ 0.80 (0.46) 1.19 (0.69) 2.32 (1.34) 2.00 (1.16) 0.69 (0.40) 2.10 (1.21) DN Thy, CD4/CD8 double-negative thymocytes; B cells, BM-resident B-cell progenitors; Eryth, BM-resident erythroid progenitors. Data are mean values ±SEM for three independent measurements generated from four pooled mice per genotype during each independent measurement.

To unequivocally demonstrate that dCK inactivation lineages compared with corresponding WT populations triggers early S-phase arrest, we performed kinetic measure- (Fig. 2 A; *, P < 0.005; **, P = 0.0001); dCTP pools were ments of DNA synthesis in vivo by measuring BrdU incor- largely unaffected (Fig. 2 A). poration into WT and dCK/ DN3b, Hardy B-C, and In contrast to dTTP pool depletions in TK1/ B cell EryA cells 1 h after intraperitoneal injection of the probe. and erythroid lineages, TK1/ DN thymocytes had normal The amount of BrdU incorporated per BrdU+ cell was re- dTTP pools, suggesting the existence of a compensatory duced significantly in all threedCK / subpopulations (Fig. 1, mechanism for dTTP production. dCMP produced by dCK E and F), a result consistent with impaired DNA synthesis. can contribute to the dTTP pool through the sequential To confirm this interpretation, we used a BrdU chase ap- actions of deoxycytidine monophosphate deaminase (DCTD) proach to analyze productive DNA synthesis in vivo (Begg and thymidylate synthase (TYMS; Fig. 1 A; Staub et al., 1988). et al., 1985; Terry and White, 2006). Mice were pulsed with To determine whether such dCK-dependent contributions BrdU, and the free circulating probe was allowed to be elimi- occur in vivo to potentially help maintain dTTP pools in the nated (i.e., chased) for 3 and 5 h before analyzing BrdU incor- absence of TK1, we measured the relative efficiencies of poration. This approach enabled detection and quantification deoxycytidine-to-thymidine conversion in WT and TK1/ of BrdU chase gaps, which are indicators of productive mice by using [13C/15N]-labeled deoxycytidine ([13C/15N]-dC). DNA synthesis (Terry and White, 2006). Although BrdU+ Both genotypes were pulsed with [13C/15N]-dC for 30 min, WT cells synthesized sufficient new DNA during the 3- and the conversion of [13C/15N]-dC to [13C/15N]-dCTP and 5-h chase points to generate typical BrdU chase gaps, and to [13C/15N]-dTTP was determined in target hemato- BrdU+ dCK/ cells did not display chase gaps, thus indi- poietic populations using liquid chromatography-tandem cating arrested DNA synthesis (Fig. 1 G). Consequently, mass spectrometry (LC-MS/MS). The [13C/15N]-dTTP to BrdU+ dCK/ cells remained stationary in early S-phase [13C/15N]-dCTP ratio was more than twofold higher in (Fig. 1 G). Moreover, few dCK/ cells entered the S-phase TK1/ DN thymocytes than in the corresponding WT sub- during the 3- and 5-h chase periods (Fig. 1 G, BrdU population (7.3:1 vs. 3.5:1; Fig. 2 B). Therefore, DN thymo- S-phase gates), further confirming that dCK inactivation cytes may compensate for the loss of the TK1-dependent perturbed cell cycle kinetics. The degree of S-phase arrest input in their dTTP pools by converting a larger fraction triggered by dCK inactivation correlated with strong up- of dCMP to dTTP (Fig. 2 A). In contrast to DN thymo- regulation of the RSR/DDR marker histone H2A.X cytes, TK1/ BM cells displayed a lower [13C/15N]-dTTP phosphorylated on Ser139 (pH2A.X) in early S-phase cells to [13C/15N]-dCTP ratio than WT BM cells (1.3:1 vs. from all three hematopoietic subpopulations (Fig. 1, H and I; 2.1:1; Fig. 2 B). There is an approximate fourfold decrease *, P < 0.0001). Although the effects of dCK inactivation in the amount of [13C/15N]-dTTP in TK1/ BM cells were highly penetrant in lymphoid and erythroid progeni- relative to WT controls (Fig. 2 B). Collectively, these find- tor populations that have short S-phases (<7 h), myeloid ings suggest that, compared with TK1/ DN thymocytes, lineage cells, which have a longer S-phase (>9 h), were TK1/ BM cells are less able to rely upon dCK to maintain significantly less affected (Fig. 1, I and J). dTTP pools. The dTTP pool depletion observed in TK1/ BM Effects of TK1 inactivation on dTTP pools, cells suggested that, similar to their dCK/ counterparts, hematopoietic development, and endogenous RS these cells would also be subjected to RS induction. BM- We next determined whether inactivation of TK1, the resident TK1/ hematopoietic populations displayed ele- other cytosolic NSP kinase expressed in mammalian cells vated pChk1 levels, whereas TK1/ DN thymocytes were (Fig. 1 A), affected hematopoietic development in a manner not affected (Fig. 2 C). These data are consistent with the similar to dCK inactivation. Unlike the severe defects in dTTP pool depletion found in BM TK1/ and with the hematopoiesis induced by dCK inactivation, lymphopoiesis normal dTTP pools in TK1/ thymocytes, respectively and erythropoiesis appeared to be significantly less affected (Fig. 2 A). Regardless of their dTTP pool status, proliferat- in the TK1/ mice (Fig. S1). Nonetheless, dTTP levels ing TK1/ hematopoietic progenitors did not display sig- were significantly decreased inTK1 / B cell and erythroid nificant changes in the cell cycle profiles (Fig. 2, D and E).

2218 Regulation of replication stress by dCK and TK1 | Austin et al. Article

Figure 2. TK1 inactivation causes only minor RS in hematopoietic cells. (A) dTTP and dCTP pools from WT (black bars) and TK1/ (gray bars) DN Thymocytes, BM- resident B cell progenitors, and BM-resident erythroid progenitors. Data are mean values ±SEM for three independent measurements. n = 4 mice/genotype/replicate; *, P < 0.005; **, P = 0.0001. (B) Concentrations (fmol per 106 cells) of [13C/15N]-dCTP (black bars) and [13C/15N]-dTTP pools (gray bars) from WT and TK1/-nucleated BM cells and DN thymocytes. Data are means ±SEM from n = 5 (WT) and n = 4 (TK1/) mice from two independent experiments. (C) Western blot detection of pChk1 in lysates from lymphoid and eryth roid progenitors. Total CHK1 protein was used as a loading control. (D) Representative example of total DNA content staining in EryA cells from WT and TK1/ cells, and (E) quantification of percentage of DN3b thymocytes, Hardy B-C cell, and EryA cells in S-phase as determined by total DNA content staining. Data are mean values ±SEM. n = 3 mice/genotype. (F) Representative example of pH2A.X detection in WT and TK1/ EryA cells. (G) Quantification of pH2A.X-positive staining in DN3b thymocytes, Hardy B-C cell, EryA cells from WT (black bars), and TK1/ (gray bars) mice. n = 4 mice/genotype. NS, P > 0.05; *, P < 0.03.

Moreover, in contrast to the highly elevated levels of pH2A. Blocking thymidine salvage alleviates the RS triggered by X found in dCK/ DN3b thymocytes and Hardy B-C dCK inactivation in BM-resident hematopoietic progenitors B cell progenitors (Fig. 1, H and I), corresponding TK1/ To further investigate the mechanistic basis of dNTP pool lymphoid cells did not up-regulate pH2A.X; only TK1/ imbalances and RS phenotypes observed in the dCK/ and EryA cells significantly up-regulated the expression of this TK1/ mice, we examined the consequences of concom- DDR marker (Fig. 2, F and G; *, P < 0.03), albeit at levels itant inactivation of these two NSP kinases. dCK/;TK1/ that were eightfold lower than those observed in dCK/ DKO hematopoietic cells should completely lack the ability EryA cells (Fig. 1 I). to salvage dNs from their extracellular environment and must

JEM Vol. 209, No. 12 2219 Figure 3. TK1 inactivation relieves the early S-phase RS in dCK/ developing B and erythroid cells. (A) Representative examples of B cell development staining of BM samples from WT, dCK/, and dCK/; TK1/ DKO mice. IgM and B220 staining of whole BM cells identify Hardy fraction A-D (B220+, IgM) and Hardy fraction E-F (B220+, IgM+) populations. Hardy fraction A-D cells are sub-gated using CD43 and CD19 expression to identify Hardy Fraction A (CD43hi, CD19), B-C (CD43hi, CD19+), and D (CD43lo, CD19+). Hardy fraction B-C cells are then analyzed for cell cycle position and pH2A.X expression. Plots are representative of n = 3 mice/genotype. (B and C) Quantification of percentage of total BM cells that are pheno- typically Hardy fraction A-D (B), and Hardy fraction B-C (C) populations from WT, dCK/, and DKO mice. Data are mean values ±SEM for n = 3/genotype; *, P < 0.04; **, P < 0.01. (D) Representative examples of pH2A.X detection in EryA cells from WT, dCK/, TK1/, and DKO mice. n = 4 mice/genotype. (E) Rep- resentative images of spleens from WT, dCK/, TK1/, and DKO mice. (F) dCTP and dTTP pool measurements from nucleated BM cells from WT, dCK/, TK1/, DKO mice. Data are mean values ±SEM. n = 4 mice/genotype. *, P < 0.001.

rely exclusively on the de novo pathway to produce mice. Instead inactivation of TK1 significantly rescued the dNTPs for DNA replication and repair. DKO mice were hematopoietic development in the dCK/;TK1/ DKO born at sub-Mendelian frequencies (hazard ratio 1.85) and mice. Thus, relative to dCK/ cells, DKO B cell progeni- weighed 30% less than WT, TK1/, and dCK/ litter- tors showed greatly improved differentiation after the Hardy mates at 5 wk of age (n = 7/genotype; P < 0.001), indicat- B-C stage (Fig. 3, A-C). Moreover, the cell cycle profile ing that the NSP is important to support animal growth. and pH2A.X expression of DKO Hardy B-C cells ap- However, combined inactivation of dCK and TK1 did not peared normal (Fig. 3 A). In the erythroid lineage, dual in- further aggravate the defects observed in the dCK single-KO activation of dCK and TK1, partially reduced the relative

2220 Regulation of replication stress by dCK and TK1 | Austin et al. Article

abundance of EryA cells in the BM (Fig. S1), but DKO EryA allosteric inhibition of RNR-mediated synthesis of deoxycyti- cells still displayed an abnormal cell cycle profile with up- dine diphosphate (dCDP), the direct precursor of dCTP regulated pH2A.X expression relative to WT and TK1/ (Reichard, 1988). It is conceivable that the RS experienced by cells (Fig. 3 D). Nonetheless, these cells were in fact relieved dCK/ hematopoietic cells in vivo can be attributed to con- of the early S-phase arrest phenotype characteristic of their comitant inhibition of both salvage and de novo pathways for dCK/ counterparts (Fig. 3 D). Intriguingly, pH2A.X up- dCTP production, therefore creating conditions of synthetic regulation in DKO EryA cells occurred in mid/late S-phase sickness/lethality (SSL). As shown in Fig. 5 A, hematopoietic rather than at the G1/S border corresponding to early stages progenitors in BM and thymus are exposed to endogenous of DNA synthesis that resulted from dCK inactivation alone concentrations of thymidine that are significantly higher than (Fig. 3 D). Altogether, erythropoiesis itself in DKO mice thymidine levels in other tissues and in plasma. Such high appeared significantly normalized, as further indicated by levels of endogenous thymidine may increase dTTP pools via the alleviation of the abnormal extramedullary erythropoiesis TK1-mediated salvage. Decreased dTTP production caused manifested as splenomegaly in the dCK/ mice (Fig. 3 E). by TK1 inactivation would relieve the inhibition of RNR’s To determine if the observation that TK1 inactivation ability to reduce CDP to dCDP in proliferating dCK/ cells rescued the early S-phase arrest in B cell and erythroid pre- residing in BM and thymus. The corresponding increase in cursors in dCK/ mice were reflected in biochemical resto- the output of the de novo pathway would therefore help ration, we investigated whether dNTP pools were also restore dCTP pools depleted by dCK inactivation. restored to normal levels. This was, indeed, the case (Fig. 3 F; To test the validity of this model we reasoned that if exces- *, P < 0.001); the dCTP pool in DKO BM cells was statisti- sive thymidine salvage in thymus and BM was indeed respon- cally indistinguishable from that of WT BM cells. In contrast sible for the dCK/ phenotype, then removal of dCK/ to the dCTP pools, dTTP pools from DKO BM cells were hematopoietic cells from their thymidine-rich in vivo environ- not restored to WT levels and remained at levels comparable ment should relieve their RS phenotype. To test this prediction, to those of TK1/ BM cells. we modeled T cell development in vitro using the OP9-DL1 co-culture system (Taghon et al., 2005). Importantly, LC-MS/ TK1 inactivation rescues the development of dCK/ T cells MS measurements showed that the OP9-DL1 culture media Similar to its effects on the development of dCK/ B cells, used in these experiments contained submicromolar amounts TK1 inactivation also rescued dCK/ T cell development. of thymidine that are comparable to plasma levels (Fig. 5 B), DKO thymi were significantly larger than those from the and are much lower than those measured in thymus and BM dCK/ mice because of increased overall cellularity (Fig. 4, (Fig. 5 A). DN3a thymocytes (CD4, CD8, CD44, CD25hi, A and B; *, P < 0.001). In DKO mice, the eightfold increase and CD27lo) from WT, dCK/, TK1/, and DKO mice in thymic cellularity over dCK/ thymi was accompanied by a were labeled with a fluorescent proliferation dye (CellTrace striking normalization of a key event in thymic T cell develop- Violet; CTV), cultured on the OP9-DL1 stroma, and then ment, specifically the transition from the DN stage to the CD4/ assessed for cell division four days later by flow cytometric CD8 double-positive (DP) stage (Fig. 4 C). Thus, in contrast to analysis of CTV dye dilution (Quah and Parish, 2012). In the dCK/ thymocytes, which experience a severe block at the contrast to their in vivo phenotype, dCK/ DN3a thymo- DN to DP transition, DKO thymocytes displayed a DN to DP cytes proliferated robustly in the in vitro co-culture system distribution that was nearly indistinguishable from that of their that contained submicromolar levels of thymidine; dCK/ WT and TK1/ counterparts (Fig. 4 C). The phenotypic res- cells divided up to 8 times, equivalent to the rate of cell divi- cue likely reflected the restoration of dCTP pools in the DKO sion observed for WT, TK1/, and DKO thymocytes over DN thymocytes (Fig. 4 D; *, P < 0.01). The DKO DN thymo- the course of the in vitro study (Fig. 5 C). These findings in- cytes also expressed significantly reduced levels of the RSR ac- dicated that salvage of endogenous thymidine in vivo con- tivation markers pChk1 and CHK2 kinase phosphorylated on tributes to the induction of S-phase arrest and RS in dCK/ Thr68 (pChk2) compared with the corresponding dCK/ hematopoietic progenitors. population (Fig. 4 E). In addition, both the early S-phase arrest To confirm the hypothesis that TK1-dependent -thymi and pH2.AX induction were significantly reduced in DKO cells dine salvage mediates S-phase arrest and RS in dCK/ relative to dCK/ DN3b thymocytes (Fig. 4 F). Collectively, hematopoietic progenitors, we titrated thymidine into the these findings indicated that elimination of TK1 activity drasti- DN3a/OP9-DL1 co-cultures and determined how this cally lowered the levels of RS in DKO versus dCK/ thymo- affected the rate of cell proliferation. The addition of 20 µM cytes, an effect explained by normalization of dCTP pools and thymidine to the co-culture medium significantly blocked reflected by the dramatic differences between T cell develop- the proliferation of dCK/ cells without affecting WT cell ment in the DKO versus dCK/ mice (Fig. 4 A and Fig. S1). division (Fig. 5 D, top row). Further increasing the thymidine concentration from 20 to 100 µM completely and specifically Endogenous thymidine plays a critical role in the induction blocked the proliferation of dCK/ cells (Fig. 5 D, bottom of a RS phenotype in dCK/ thymocytes row). As expected, TK1/ and DKO cells were unaffected In in vitro studies, excessive levels of intracellular dTTP by thymidine at either concentration (Fig. 5 D), because of have been shown to negatively regulate dCTP pools through their inability to salvage this deoxyribonucleoside.

JEM Vol. 209, No. 12 2221 Figure 4. TK1 inactivation normalizes the development of dCK/ T cells. (A and B) Gross thymus size (A), and total viable thymocytes recovered (B), from WT, dCK/, TK1/, and DKO mice. Data are mean values ±SEM. n = 5 mice/genotype; *, P < 0.001. (C) Repre- sentative example of CD4 and CD8 staining of total thymocytes from WT, dCK/, TK1/, and DKO mice. Percentages of DN thymocytes (bottom left gate) and DP thymocytes (top right gate) are indicated. (D) Measurements of dCTP pools in DN thymocytes isolated from WT, dCK/, TK1/, and DKO mice. Data are mean values ±SEM. n = 2 mice/genotype; *, P < 0.01. (E) Western blot detection of pChk1 and Chk2 phosphorylated on Thr68 (pChk2) in DN thymocytes. Actin was used as a loading control. (F) Representative examples of pH2A. X detection in DN3b thymocytes from WT, dCK/, TK1/, and DKO mice.

To verify that the thymidine-induced proliferation block via a TK1-independent mechanism, we titrated the RNR in- was caused by RS induction, we assayed pH2A.X activation hibitor hydroxyurea into the DN3b/OP9-DL1 co-cultures, in DN3b/OP9-DL1 co-cultures in response to thymidine and then measured RS induction 12 h later. WT, dCK/, concentrations varying from 10 to 200 µM. dCK/ thymo- TK1/, and DKO cells were equally susceptible to hydroxy- cytes strongly induced pH2A.X after exposure to thymidine urea concentrations ranging from 10–200 µM (Fig. 5 F), sug- concentrations as low as 10 µM. pH2A.X expression by gesting that dCK plays a specific role in preventing RS induced dCK/ cells peaked at 20 µM thymidine (Fig. 5 E); higher by thymidine. In conclusion, data from the OP9-DL1 co-cul- thymidine concentrations triggered massive cell death in the ture in vitro T cell development system showed that dCK- / dCK DN3b/OP9-DL1 co-cultures (percent sub-G1 cells deficient cells were hypersensitive to RS induced by low >80% at 50 µM or greater thymidine concentrations; n = 3 amounts of thymidine through a TK1-dependent mechanism. independent experiments; unpublished data). WT cells were 10-fold more resistant to thymidine than dCK/ cells; DISCUSSION TK1/ and DKO cells were completely resistant to thymi- Using genetic models of individual and combined deficien- dine at all tested concentrations (Fig. 5 E). cies in the cytosolic dN kinases dCK and TK1, we demon- Lastly, to determine whether dCK/ cells were intrinsi- strate that the dN salvage pathway contributes both to induction cally more susceptible to RS induced by nucleotide deprivation and to prevention of RS during hematopoietic development.

2222 Regulation of replication stress by dCK and TK1 | Austin et al. Article

Figure 5. Thymidine induces RS in cultured dCK/ thymocytes. (A) Relative thymidine abundance in various tissues from C57BL/6 mice as determined by LC-MS/MS. Concentrations given in mol/g of whole tissue. Data are means ±SEM; n = 5/tissue type. (B) Concentrations of thymidine (in M) from C57BL/6 plasma and from standard OP9-DL1 culture medium, as determined by LC-MS/MS. Data are means ±SEM; Plasma, n = 7; Media, n = 3. (C) Representative Cell- Trace Violet (CTV) dye dilution curves from WT, dCK/, TK1/, and DKO DN3a thymocytes cultured on OP9-DL1 stroma for 4 d without thymidine supplementation in the medium. Red numbers above the distinct CTV peaks reflect the total number of com- pleted cell divisions executed by the thymocyte populations. WT, dCK/, and TK1/, n = 4; DKO, n = 1. (D) CTV dye dilution curves after 4 d of culturing in the presence of 20 and 100 µM thymidine added to the culture medium. WT, dCK/, and TK1/, n = 4; DKO, n = 1. (E) Percent of live cells that are pH2A.X positive in WT (black circles), dCK/ (blue squares), TK1/ (red up triangles), and DKO (green down triangles) DN3b thymocytes after 48 h of stimulation in increasing concentrations of thymidine. WT, dCK/, and TK1/, n = 2; DKO, n = 1. *, Cessation of measurements of pH2A.X levels caused by

massive cell death (>80% sub-G1 staining; not depicted) induced by exposure of dCK/ cells to concentrations of thymidine equal to or greater than 50 µM. (F) pH2A.X expression in WT, dCK/, TK1/, and DKO DN3b thymocytes after 12 h of exposure to increasing concentrations of hydroxyurea 36 h after plating on OP9-DL1 stroma. WT, dCK/, and TK1/, n = 2; DKO, n = 1.

We also show that endogenous thymidine itself is a highly ac- conditions. Our data challenge this assumption by establishing tive metabolite that induces RS in vivo, an observation remi- thymidine block as a metabolic phenomenon that normally niscent of a widely used in vitro experimental approach for occurs in BM and thymus during hematopoiesis. We show that, synchronizing cells in cell cycle known as the “thymidine in the absence of dCK activity, physiological concentrations block” (Xeros, 1962). In the course of a typical thymidine block of thymidine in these tissues are sufficient to induce dCTP experiment, cultured cells exposed to high concentrations of deprivation and severe RS in proliferating T cell, B cell, and thymidine undergo S-phase arrest through a TK1-dependent erythroid precursors. By enabling these cells to salvage deoxy- mechanism (Gagou et al., 2010). It is through the kinase action cytidine, dCK replenishes dCTP pools depleted by dTTP pro- of TK1 that exogenously added thymidine is trapped in cells, duced from thymidine via TK1, thereby exerting an important thus enabling its conversion to dTTP in the cytosol. dTTP is role in hematopoiesis. The two cytosolic dN salvage kinases not only a precursor of DNA but also a strong allosteric inhibi- thus play paradoxically opposing roles; TK1 in the absence of tor of RNR’s affinity for its pyrimidine ribonucleotide sub- dCK induces RS, whereas dCK functions to prevent RS in strates CDP and uridine diphosphate (UDP; Fairman et al., TK1-expressing cells exposed to elevated levels of endogenous 2011). dTTP-mediated inhibition of CDP conversion to dCDP thymidine (Fig. 6, A and B). Importantly, the degree of depen- depletes dCTP pools, thus causing S-phase arrest (Bjursell and dence on the NSP kinases to avoid endogenous RS varied Reichard, 1973; Larsson et al., 2004). The thymidine block ap- across different hematopoietic lineages, with the erythroid lin- proach has been thought of traditionally as an in vitro “labora- eage being most affected, followed by T and B cell progenitors tory tool” (Gentry, 1992) with little, if any, relevance to in vivo and then by myeloid precursors (Fig. 1 J and Fig. 6 B).

JEM Vol. 209, No. 12 2223 The causal relationship between TK1 activity, inhibition DNA during DNA repair (Hu et al., 2012). Second, by of de novo dCTP production, and dependency on dCK to opposing the activity of cytosolic nucleotidases, which avoid thymidine-induced endogenous RS inevitably leads to dephosphorylate dTMP, TK1 activity will prevent the loss of the question of why rapidly dividing hematopoietic cells en- thymidine containing deoxyribonucleotides from dividing gage in what at first sight would appear to be excessive salvage cells (He and Skog, 2002). Third, by increasing cytosolic dTTP of thymidine. Further adding to the puzzle of why hemato- pools, TK1 would promote the allosteric inhibition of RNR’s poietic cells express high levels of TK1 is that our data show ability to reduce not only CDP but also UDP (Fairman et al., that hematopoiesis appears to be much less severely affected by 2011; Fig. 6 B), therefore preventing excessive accumulation TK1 inactivation than by the loss of dCK activity. Although of dUTP caused by high RNR activity in proliferating cells. additional studies are needed to fully elucidate the biological If confirmed by future studies, this hypothesis may explain significance of TK1 by further analyses of the consequences why the dTTP-mediated allosteric regulation of RNR’s sub- of its inactivation on hematopoiesis, our preliminary data in- strate specificity is mostly conserved among members of the dicate that TK1 may play an important role in regulating the three classes of RNRs, despite wide differences in their pri- dUTP/dTTP balance; we observe an approximately fourfold mary and quaternary structures (Reichard, 1993). increase in dUTP/dTTP ratio in TK1/ BM cells relative to WT BM cells (unpublished data). A proper dUTP/dTTP The role of the NSP in nonhematopoietic tissues ratio is needed to reduce misincorporation of dUTP into DNA Our findings indicate that the de novo synthesis pathway can (Melnyk et al., 1999), thereby preventing the formation of maintain dNTP pools at levels that are sufficient to support U:A pairs. Excessive occurrences of such pairs may have cyto- most hematopoietic proliferation. However, the observation toxic effects (Hagen et al., 2006) and would require the inter- that DKO mice are born at sub-Mendelian frequencies and that vention of uracil-DNA glycosylases and of base excision surviving animals display growth retardation indicates that repair mechanisms to prevent induction of RS. TK1 activity the NSP is important in supporting overall growth. It is pos- could promote the maintenance of a low dUTP/dTTP ratio sible that the growth defects affecting the DKO mice may in three ways. First, through preferential phosphorylation of result from increased stress forced upon other biosynthetic thymidine versus deoxyuridine caused by a 20-fold difference pathways in the absence of the NSP. The utilization of dN in affinity in favor of thymidine (Munch-Petersen et al., 1991), salvage for dNTP synthesis requires significantly less ATP and TK1 is likely to generate significantly more dTMP than NADPH than does de novo synthesis (Lunt and Vander dUMP. In turn, this would increase the substrate availability Heiden, 2011). For example, de novo synthesis of 1 mol of for thymidylate kinase, an enzyme recently shown to play an either dCTP or dTTP starting from glucose, glutamine, aspar- important role in preventing dUTP incorporation into the tate, and bicarbonate requires 6 moles of ATP each, while the

Figure 6. Deoxyribonucleoside salvage kinases induce and resolve RS during hematopoiesis. (A) Under normal conditions, the RNR complex reduces purine ribonucleotide diphosphates (ADP and GDP) and pyrimidine ribonucleotide diphosphates (CDP and UDP) to contribute to dNTP pools (dATP, dGTP, dTTP, and dCTP). Although in hematopoietic cells RNR appears to be solely responsible for producing purine dNTP pools, the majority of dTTP pools are produced from salvaged thymidine, which is present in abundant amounts in thymus and BM. dCK may also contribute to dTTP pools as shown in Fig. 1 A and Fig. 2 C. Elevated dTTP levels prevent RNR from reducing CDP to dCDP and possibly UDP to dUDP via allosteric inhibition. To maintain dCTP pools, rapidly dividing hematopoietic cells rely on deoxycytidine salvage via dCK. (B) Graphical representation of the source (D-de novo, S-salvage) and size (height of D or S) of dCTP pools in various hematopoietic lineages. In the absence of dCK activity (dCK/ column), dCTP pools become insufficient, leading to severe RS (++++) and DNA synthesis arrest in early S-phase in T cell, B cell, and erythroid cell precursors. In the absence of TK1 activity (TK1/ column), dCTP pools are unaffected and only mild RS (+) occurs in late S-phase in erythroid precursors. The mild RS may be caused by an imbalanced dUTP/dTTP ratio in the absence of TK1. When both dCK and TK1 are inactivated (DKO column), de novo dCDP production is de-repressed and, subse- quently, dCTP pools are restored to WT levels. DKO thymocytes have measurable, but overall mild, levels of RS (+) in early S-phase. The absence of NSP is well tolerated by B cell precursors, but it results in severe late S-phase RS (+++) in erythroid precursors.

2224 Regulation of replication stress by dCK and TK1 | Austin et al. Article

same amount of dCTP or dTTP can be generated via salvage overload, and ultimately cell death, in hyperproliferative dis- at a cost of 3 moles of ATP (Voet and Voet, 2004). ATP and eases such as cancer. Thymidine therapy has been attempted NADPH savings afforded by using pyrimidine salvage instead in human malignancies, based on the ability of this nucleoside of de novo synthesis could then be spent by cells to generate to induce S-phase arrest of cultured cancer cells by inhibiting other essential biomass, such as proteins and phospholipids. their de novo dCTP synthesis. Although well tolerated in An alternative explanation for the growth defects in the DKO patients, thymidine had limited efficacy when used as single- mice involves mitochondrial dysfunction due to imbalances agent therapy in acute lymphoid and myeloid leukemias in the dNTP pools in this organelle. Defects in mitochondrial (Kufe et al., 1980). By demonstrating that dCK, which is fre- dNTP synthesis have been shown to cause severe develop- quently overexpressed in hematological malignancies, signifi- mental abnormalities (Eriksson and Wang, 2008; Zhou et al., cantly contributes to dCTP pools in vivo, the current study 2008; Zhou et al., 2010; González-Vioque et al., 2011). Since may have revealed an important mechanism of resistance to in cycling cells mitochondrial dNTP pools correlate linearly thymidine and thus explain the failure of thymidine in initial with the cytoplasmic pools (Gandhi and Samuels, 2011), im- clinical trials. Combining thymidine with a small molecule balances in cytosolic dNTP pools resulting from defects in dCK inhibitor (dCKi) would induce SSL by blocking both the NSP may impair mitochondrial DNA synthesis and thus dCTP-producing pathways. Whether an adequate therapeu- trigger mitochondrial stress manifested as embryonic lethality tic window exists for thymidine/dCKi SSL will be deter- and neonatal growth retardation. mined in subsequent studies.

Nucleotide deficiency, cellular transformation, and the relationship between the NSP and the RSR pathway MATERIALS AND METHODS Nucleotide deficiency has recently been shown to promote Animals. Mice were bred and housed under specific pathogen–free conditions and were treated in accordance with the UCLA Animal Research cellular transformation (Bester et al., 2011). The study by Committee protocol guidelines. The dCK/ mice were generated and bred Bester et al. described how overexpression either of viral as previously described and backcrossed to C57BL/6 for n = 7 generations E-proteins or of mammalian Cyclin E in cultured human ke- (Toy et al., 2010). TK1/ mice were rederived from cryopreserved embry- ratinocytes overrides normal cell cycle regulation and triggers onic stem cells stored at the UC Davis Repository (stock VG18248) and premature S-phase entry, which in turn causes RSR pathway genotyped as previously described (Dobrovolsky et al., 2003). dCK/TK1 +/ +/ activation and subsequent mutation induction. Presumably, DKO mice were generated by first intercrossing dCK mice to TK1 mice to generate dCK+/ TK1+/ progeny. dCK+/ TK1+/ mice were this chain of events is initiated by the inability of E-protein/ then intercrossed together to generate dCK/TK1 DKO mice at potential Cyclin E–overexpressing keratinocytes to generate sufficient 1:16 frequency. Age-matched (5–12-wk-old) WT, dCK/, TK1/, and dNTP pools to support DNA replication and repair. The nu- DKO littermates were used for all experiments except for the endogenous cleotide deficiency observed here in our in vivo genetic sys- thymidine determinations from whole tissue. Tissues for endogenous thymi- tems resembles that induced by Bester et al. (2011) in their dine measurements were extracted from 6-wk-old C57BL/6 mice purchased cell culture system. However, despite the severe nucleotide from the UCLA Radiation Oncology breeding colony. deficiency and strong activation of the RSR pathway ob- Tissue preparation. Single-cell suspensions were prepared from BM by / served in the dCK hematopoietic lineages, we did not de- flushing femur, tibia, and humerus bones with DMEM supplemented with / tect increased rates of cancer in the dCK mice when these 2% FBS using 25-G needles, followed by 70-µm filtration. Whole BM was animals were followed for 12 mo after birth. It is possible that depleted of mature red blood cells by overlaying cell suspensions on a solu- onset of nucleotide deficiency-induced cancer in our model tion of 16% iodixanol (D1556; Sigma-Aldrich), 0.63% NaCl, 10 mM Hepes, will occur at an even older age. It is also conceivable that the pH 7.4, and 0.1% NaN3, followed by centrifugation at 900 g for 15 min. protective effects of the RSR pathway in dCK/ mice are The mononuclear cell–containing supernatant was transferred and washed twice before antibody staining or dNTP extraction. Single-cell suspensions sufficient to eliminate potentially oncogenic cells, whereas a of thymus were prepared by mechanical dissociation using frosted glass slides yet to be defined aspect of the keratinocyte system permits in DMEM supplemented with 2% FBS and 50 µg/ml DNaseI (Roche) and the propagation of cells carrying fixed DNA mutations. Fu- passed through 70-µm sterile filters. CD4/CD8 thymocytes were purified ture studies will be required to determine the consequences from whole thymocyte suspensions using combined CD4 and CD8 negative of concomitant inactivation of the NSP and of key compo- selection kits (Invitrogen). nents of the RSR pathway. In this context, it is important to Immunophenotyping antibodies. The following antibodies from eBio- note the similarities between the defects in erythropoiesis ob- science were used for B cell, erythroid cell, and myeloid cell phenotyping served in the NSP-deficient mice and defects in this hemato- from whole BM: B220 (clone RA3-6B2) PE-Cy7 and APC-eFluor780; poietic lineage documented in mouse models of ATR-Seckel IgM (clone II/41) FITC, PerCP-eFluor710; CD43 (clone eBioR2/60) PE; (Murga et al., 2009) and of Chk1 haploinsufficiency (Boles CD19 (clone 1D3) APC and PE-Cy7; Ter119 (clone TER-119) PerCP- et al., 2010). Cy5.5 and PE-Cy7; CD71 (clone R17217) APC and PE; and CD11b (clone M1/70) APC-eFluor780. The following antibodies from eBioscience were used for thymocyte phenotyping: CD4 (clone L3T4) FITC, PE, Alexa Fluor Therapeutic implications 700, and PE-Cy7; CD8a (clone 53–6.7) PE, PE-Cy7, and PerCP-eFluor710; The demonstration of a functional link between endogenous CD25 (clone PC61.5) APC and PE-Cy7; CD44 (clone IM7) APC- thymidine, dN kinases, nucleotide deficiency, and RS induc- eFluor780; CD27 (clone LG.7F9) PE; and CD45 (clone 30-F11) PE-Cy7. tion highlights a potential therapeutic strategy to induce RS Hematopoietic stem cell phenotyping from whole BM: Lineage cocktail PE,

JEM Vol. 209, No. 12 2225 CD150 PE-Cy5 (clone TC15-12F12.2), (BioLegend); CD127 APC- SVC100H SpeedVac Concentrator (Savant). Dry pellets were resuspended eFluor780 (clone A7R34), Flk2/Flt3 Biotin (clone A2F10), Sca-1 PerCP- in 100 µl ddH2O, vortexed and centrifuged for 15 min at 17,000 g at 4°C to Cy5.5 (clone D7), c-Kit PE-Cy7 (clone 2B8), CD34 eFluor660 (clone clear insoluble debris. 5 µl of concentrated lysate was used in a 25-µl reaction RAM34), CD16/32 Alexa Fluor 700 (clone 93; eBioscience); streptavidin volume. Reactions were performed for 2 h according to previously described PE-Alexa Fluor 610 (Invitrogen). protocols (Mathews and Wheeler, 2009).

Flow cytometry analyses. All flow cytometry data were acquired on four- In vivo measurements of dCK-dependent incorporation of and five-laser LSRII cytometers (BD) for analysis, and FACS purification of [13C/15N]-deoxycytidine into dCTP and dTTP pools. Mice were in- cells was performed on four-laser BD FACSAriaII cell sorters running BD jected intraperitoneally with 200 µl of 2.5 mM uniformly labeled [13C/15N]- FACSDiva6 software (BD). All cytometry data were analyzed using FlowJo deoxycytidine (Cambridge Isotopes). Mice were euthanized 30 min after software (Tree Star). injection to harvest BM cells and DN thymocytes. Cells were washed with ice cold PBS/2% FBS, and then extracted with 60% methanol overnight at DNA content staining and intracellular detection of BrdU and 20°C, evaporated, and the dry pellets were resuspended in 100–500 µl of pH2A.X. BrdU (1 mg/mouse) was administered by intraperitoneal injec- 5 mM hexylamine. dTTP/dCTP concentrations were then determined by tion. Cells were collected, antibody-stained for surface antigens, and then LC-MS/MS using an Agilent 6460 Triple Quad LC/MS system. The sta- processed for intracellular detection of BrdU and pH2A.X using a BrdU- tionary phase was a Phenomenex Gemini-NX C18 column (2.0 × 100 mM) FITC kit staining protocol and reagents (BD). Cells were then stained with with 5 mM hexylamine in ddH2O and acetonitrile as the mobile phases. BrdU-FITC and/or pH2A.X antibodies conjugated to FITC (Millipore; clone JBW301) or Alexa Fluor 647 (BD; clone N1-431). Total DNA con- Whole-tissue nucleoside concentration measurements. Solid tissues tent was assessed by staining with DAPI (Roche) at 1 µg/ml final concentra- were homogenized at 4°C in acetonitrile/methanol (9:1) using the Bullet- tion in PBS containing 2% FBS. Blender Tissue Homogenizer (NextAdvance) according to the manufactur- er’s specifications. BM cells were flushed using ice-cold PBS, filtered into CTV labeling of thymocytes. Single-cell suspensions of thymocytes were single-cell suspensions, pelleted, and resuspended in homogenization buffer. resuspended in PBS containing 0.5% FBS at a cell density of 50 × 106 cells/ml. Whole blood was collected in heparinized tubes, and then centrifuged to 5 mM stocks of CTV dye (Invitrogen) dissolved in DMSO were diluted to separate plasma and blood cells. Both fractions were extracted directly with 50 µM in PBS/0.5% FBS and then added at a dilution of 1:10 to the cell the homogenization buffer. Extracted samples were evaporated, and the dry suspension (5 µM final concentration). Cells were mixed well, incubated at pellets were resuspended in ddH2O. Thymidine concentrations were then 37°C for 20 min, and then washed twice with 40 ml of PBS/5% FBS. Cells determined by LC-MS/MS. The stationary phase was a Thermo Fisher Hy- were then stained with fluorescent antibodies to identify DN3a thymocytes percarb column (2.1 × 100 mM) with 0.1% formic acid in ddH2O and 0.1% and purified via FACS. formic acid in acetonitrile as the mobile phases.

OP9-DL1 co-cultures. OP9 cells were purchased from American Type Statistical analyses. All statistics presented as means of biological replicates Culture Collection (CRL-2749) and transduced by retroviral infection with with standard error of the mean (±SEM). P-value significances were calcu- delta-like Ligand 1 vector (DL1), provided by J. Carlos Zúñiga-Pflücker lated using one sample Student’s t test function in GraphPad Prism 5 (Graph- (University of Toronto, Toronto, Ontario, Canada), and FACS-sorted based Pad Software). on their GFP fluorescence. OP9-DL1 cells were maintained in Minimum Essential Medium, Alpha Modification (Sigma-Aldrich) supplemented with Online supplemental material. Figure S1 shows representative examples 20% FBS, 100 U/ml penicillin, 100 µg/ml streptomycin. Thymocyte/OP9- of hematopoietic stem cell phenotyping, thymocyte phenotyping, BM B cell / DL1 co-cultures were initiated by plating 50 × 103 OP9-DL1 cells in 250 µl developmental phenotyping, BM erythroid phenotyping for WT, dCK , / of media/well in 48-well tissue culture dishes at day 1. On day 0, DN3a TK1 , and DKO mice. Online supplemental material is available at thymocytes were FACS purified and resuspended in OP9-DL1 media http://www.jem.org/cgi/content/full/jem.20121061/DC1. supplemented with 10 ng/ml Flt3-ligand (PeproTech) and 10 ng/ml IL-7 (PeproTech) at a cell density of 105 cells/ml. The single-cell suspensions We thank the flow cytometry cores of the UCLA Jonsson Comprehensive Cancer (250 µl) were then plated atop OP9-DL1 monolayers (25 × 103 cells; 5 ng/ml Center, UCLA Broad Stem Cell Research Center, and the UCLA Institute of Molecular Medicine for use of their equipment. We acknowledge Dr. Kym F. Faull and the final concentrations of Flt3-ligand and IL-7 in a 500 µl total volume). Co- UCLA Pasarow Mass Spectrometry laboratory for their assistance with LC-MS/MS cultures were incubated for 2 or 4 d. Thymocytes were harvested by forceful measurements. We acknowledge Drs. Steve Bensinger and Mike Teitell for critically pipetting, stained for developmental antigens and intracellular markers, and reading the manuscript. analyzed by flow cytometry. This work was funded by In Vivo Cellular and Molecular Imaging Centers Developmental Project Award National Institutes of Health P50 CA86306 Western blots. Purified hematopoietic cell populations were lysed in 1X (C.G. Radu, H.R. Herschman, and W.R. Austin), and the US National Cancer Institute RIPA buffer containing 1X Halt Protease/Phosphatase Inhibitor (78440; grant 5U54 CA119347 (C.G. Radu). O.N. Witte is an Investigator of the Howard Thermo Fisher Scientific); supernatants were isolated after centrifugation at Hughes Medical Institute. 17,000 g for 15 min. Lysates were mixed with 1X Laemmli-SDS loading The authors have no conflicting financial interests. buffer, boiled, electrophoresed, and transferred to nitrocellulose membranes for immunoblotting. Monoclonal rabbit anti–phospho-CHK1 (Ser345; Submitted: 18 May 2012 clone 133D3); monoclonal mouse anti-CHK1 (clone 2G1D5); and mono- Accepted: 3 October 2012 clonal rabbit anti-Phospho-CHK2 (Thr48; clone C13C1) were purchased from Cell Signaling Technology. Monoclonal mouse anti-actin (clone REFERENCES MM2/193) was purchased from Sigma-Aldrich. Arlt, M.F., A.C. Ozdemir, S.R. Birkeland, T.E. Wilson, and T.W. Glover. 2011. Hydroxyurea induces de novo copy number variants in human cells. Intracellular dNTP pool measurements. Purified hematopoietic cell Proc. Natl. Acad. Sci. USA. 108:17360–17365. http://dx.doi.org/10.1073/ populations were counted and pelleted. Pellets were then suspended in 1 ml pnas.1109272108 of ice-cold 60% methanol, vortexed for 1 min and stored overnight at -20°C. Arnér, E.S., and S. Eriksson. 1995. Mammalian deoxyribonucleoside The following day, the lysates were boiled for 3 min and then centrifuged kinases. Pharmacol. Ther. 67:155–186. http://dx.doi.org/10.1016/0163- for 15 min at 17,000 g at 4°C. Supernatants were evaporated overnight in a 7258(95)00015-9

2226 Regulation of replication stress by dCK and TK1 | Austin et al. Article

Begg, A.C., N.J. McNally, D.C. Shrieve, and H. Kärcher. 1985. A method González-Vioque, E., J. Torres-Torronteras, A.L. Andreu, and R. Martí. to measure the duration of DNA synthesis and the potential dou- 2011. Limited dCTP availability accounts for mitochondrial DNA bling time from a single sample. Cytometry. 6:620–626. http://dx.doi depletion in mitochondrial neurogastrointestinal encephalomyopathy .org/10.1002/cyto.990060618 (MNGIE). PLoS Genet. 7:e1002035. http://dx.doi.org/10.1371/journal Bester, A.C., M. Roniger, Y.S. Oren, M.M. Im, D. Sarni, M. Chaoat, .pgen.1002035 A. Bensimon, G. Zamir, D.S. Shewach, and B. Kerem. 2011. Hagen, L., J. Peña-Diaz, B. Kavli, M. Otterlei, G. Slupphaug, and H.E. Nucleotide deficiency promotes genomic instability in early stages of Krokan. 2006. Genomic uracil and human disease. Exp. Cell Res. cancer development. Cell. 145:435–446. http://dx.doi.org/10.1016/j 312:2666–2672. http://dx.doi.org/10.1016/j.yexcr.2006.06.015 .cell.2011.03.044 Halazonetis, T.D., V.G. Gorgoulis, and J. Bartek. 2008. An oncogene- Bjursell, G., and P. Reichard. 1973. Effects of thymidine on deoxyribo- induced DNA damage model for cancer development. Science. 319:1352– nucleoside triphosphate pools and deoxyribonucleic acid synthesis in 1355. http://dx.doi.org/10.1126/science.1140735 Chinese hamster ovary cells. J. Biol. Chem. 248:3904–3909. Hardy, R.R., P.W. Kincade, and K. Dorshkind. 2007. The protean nature Boles, N.C., S. Peddibhotla, A.J. Chen, M.A. Goodell, and J.M. Rosen. 2010. of cells in the B lymphocyte lineage. Immunity. 26:703–714. http:// Chk1 haploinsufficiency results in anemia and defective erythropoiesis. dx.doi.org/10.1016/j.immuni.2007.05.013 PLoS ONE. 5:e8581. http://dx.doi.org/10.1371/journal.pone.0008581 He, Q., and S. Skog. 2002. Involvement of a substrate cycle between thy- Branzei, D., and M. Foiani. 2008. Regulation of DNA repair through- midine and thymidylate in the regulation of DNA precursor pool in out the cell cycle. Nat. Rev. Mol. Cell Biol. 9:297–308. http://dx.doi ehrlich ascites tumour. Cell. Physiol. Biochem. 12:305–314. http://dx.doi .org/10.1038/nrm2351 .org/10.1159/000067900 Chen, Y.L., S. Eriksson, and Z.F. Chang. 2010. Regulation and func- Hu, C.M., M.T. Yeh, N. Tsao, C.W. Chen, Q.Z. Gao, C.Y. Chang, M.H. tional contribution of thymidine kinase 1 in repair of DNA damage. Lee, J.M. Fang, S.Y. Sheu, C.J. Lin, et al. 2012. Tumor cells require J. Biol. Chem. 285:27327–27335. http://dx.doi.org/10.1074/jbc.M110 thymidylate kinase to prevent dUTP incorporation during DNA repair. .137042 Cancer Cell. 22:36–50. http://dx.doi.org/10.1016/j.ccr.2012.04.038 Choi, O., D.A. Heathcote, K.K. Ho, P.J. Müller, H. Ghani, E.W. Lam, Kufe, D.W., P. Beardsley, D. Karp, L. Parker, A. Rosowsky, G. Canellos, P.G. Ashton-Rickardt, and S. Rutschmann. 2012. A deficiency in and E. Frei III. 1980. High-dose thymidine infusions in patients with nucleoside salvage impairs murine lymphocyte development, homeo- leukemia and lymphoma. Blood. 55:580–589. stasis, and survival. J. Immunol. 188:3920–3927. http://dx.doi.org/10 Larsson, K.M., A. Jordan, R. Eliasson, P. Reichard, D.T. Logan, and P. .4049/jimmunol.1102587 Nordlund. 2004. Structural mechanism of allosteric substrate specificity D’Angiolella, V., V. Donato, F.M. Forrester, Y.T. Jeong, C. Pellacani, Y. regulation in a ribonucleotide reductase. Nat. Struct. Mol. Biol. 11:1142– Kudo, A. Saraf, L. Florens, M.P. Washburn, and M. Pagano. 2012. 1149. http://dx.doi.org/10.1038/nsmb838 Liu, Y., R. Pop, C. Sadegh, C. Brugnara, V.H. Haase, and M. Socolovsky. Cyclin F-mediated degradation of ribonucleotide reductase M2 controls 2006. Suppression of Fas-FasL coexpression by erythropoietin mediates genome integrity and DNA repair. Cell. 149:1023–1034. http://dx.doi erythroblast expansion during the erythropoietic stress response in vivo. .org/10.1016/j.cell.2012.03.043 Blood. 108:123–133. http://dx.doi.org/10.1182/blood-2005-11-4458 Dobrovolsky, V.N., T. Bucci, R.H. Heflich, J. Desjardins, and F.C. Lunt, S.Y., and M.G. Vander Heiden. 2011. Aerobic glycolysis: meeting the Richardson. 2003. Mice deficient for cytosolic thymidine kinase gene metabolic requirements of cell proliferation. Annu. Rev. Cell Dev. Biol. develop fatal kidney disease. Mol. Genet. Metab. 78:1–10. http://dx.doi 27:441–464. http://dx.doi.org/10.1146/annurev-cellbio-092910-154237 .org/10.1016/S1096-7192(02)00224-X Mathews, C.K., and L.J. Wheeler. 2009. Measuring DNA precursor pools Dobrovolsky, V.N., L.J. McGarrity, L.S. VonTungeln, R.A. Mittelstaedt, in mitochondria. Methods Mol. Biol. 554:371–381. http://dx.doi.org/ S.M. Morris, F.A. Beland, and R.H. Heflich. 2005. Micronucleated 10.1007/978-1-59745-521-3_22 erythrocyte frequency in control and azidothymidine-treated Tk+/+, Matsuoka, S., B.A. Ballif, A. Smogorzewska, E.R. McDonald III, K.E.   Tk+/- and Tk / mice. Mutat. Res. 570:227–235. http://dx.doi Hurov, J. Luo, C.E. Bakalarski, Z. Zhao, N. Solimini, Y. Lerenthal, .org/10.1016/j.mrfmmm.2004.11.006 et al. 2007. ATM and ATR substrate analysis reveals extensive protein Eriksson, S., and L. Wang. 2008. Molecular mechanisms of mitochondrial networks responsive to DNA damage. Science. 316:1160–1166. http:// DNA depletion diseases caused by deficiencies in enzymes in purine and dx.doi.org/10.1126/science.1140321 pyrimidine metabolism. Nucleosides Nucleotides Nucleic Acids. 27:800–808. Melnyk, S., M. Pogribna, B.J. Miller, A.G. Basnakian, I.P. Pogribny, and http://dx.doi.org/10.1080/15257770802146197 S.J. James. 1999. Uracil misincorporation, DNA strand breaks, and gene Evans, D.R., and H.I. Guy. 2004. Mammalian pyrimidine biosynthesis: amplification are associated with tumorigenic cell transformation in fo- fresh insights into an ancient pathway. J. Biol. Chem. 279:33035–33038. late deficient/repleted Chinese hamster ovary cells.Cancer Lett. 146:35– http://dx.doi.org/10.1074/jbc.R400007200 44. http://dx.doi.org/10.1016/S0304-3835(99)00213-X Fairman, J.W., S.R. Wijerathna, M.F. Ahmad, H. Xu, R. Nakano, S. Jha, Munch-Petersen, B., L. Cloos, G. Tyrsted, and S. Eriksson. 1991. Diverging J. Prendergast, R.M. Welin, S. Flodin, A. Roos, et al. 2011. Structural substrate specificity of pure human thymidine kinases 1 and 2 against basis for allosteric regulation of human ribonucleotide reductase by antiviral dideoxynucleosides. J. Biol. Chem. 266:9032–9038. nucleotide-induced oligomerization. Nat. Struct. Mol. Biol. 18:316–322. Murga, M., S. Bunting, M.F. Montaña, R. Soria, F. Mulero, M. Cañamero, http://dx.doi.org/10.1038/nsmb.2007 Y. Lee, P.J. McKinnon, A. Nussenzweig, and O. Fernandez-Capetillo. Fustin, J.M., M. Doi, H. Yamada, R. Komatsu, S. Shimba, and H. 2009. A mouse model of ATR-Seckel shows embryonic replicative Okamura. 2012. Rhythmic nucleotide synthesis in the liver: tempo- stress and accelerated aging. Nat. Genet. 41:891–898. http://dx.doi.org/ ral segregation of metabolites. Cell Rep. 1:341–349. http://dx.doi.org/ 10.1038/ng.420 10.1016/j.celrep.2012.03.001 Niida, H., M. Shimada, H. Murakami, and M. Nakanishi. 2010. Mechanisms Gagou, M.E., P. Zuazua-Villar, and M. Meuth. 2010. Enhanced H2AX of dNTP supply that play an essential role in maintaining genome in- phosphorylation, DNA replication fork arrest, and cell death in the ab- tegrity in eukaryotic cells. Cancer Sci. 101:2505–2509. http://dx.doi sence of Chk1. Mol. Biol. Cell. 21:739–752. http://dx.doi.org/10.1091/ .org/10.1111/j.1349-7006.2010.01719.x mbc.E09-07-0618 Poli, J., O. Tsaponina, L. Crabbé, A. Keszthelyi, V. Pantesco, A. Chabes, A. Gandhi, V.V., and D.C. Samuels. 2011. A review comparing deoxyri- Lengronne, and P. Pasero. 2012. dNTP pools determine fork progres- bonucleoside triphosphate (dNTP) concentrations in the mitochon- sion and origin usage under replication stress. EMBO J. 31:883–894. drial and cytoplasmic compartments of normal and transformed cells. http://dx.doi.org/10.1038/emboj.2011.470 Nucleosides Nucleotides Nucleic Acids. 30:317–339. http://dx.doi.org/ Pontarin, G., P. Ferraro, L. Bee, P. Reichard, and V. Bianchi. 2012. Mammalian 10.1080/15257770.2011.586955 ribonucleotide reductase subunit p53R2 is required for mitochondrial Gentry, G.A. 1992. Viral thymidine kinases and their relatives. Pharmacol. DNA replication and DNA repair in quiescent cells. Proc. Natl. Acad. Sci. Ther. 54:319–355. http://dx.doi.org/10.1016/0163-7258(92)90006-L USA. 109:13302–13307. http://dx.doi.org/10.1073/pnas.1211289109

JEM Vol. 209, No. 12 2227 Quah, B.J., and C.R. Parish. 2012. New and improved methods for mea- beta- and gammadelta-selected pre-T cells in the adult mouse thymus. suring lymphocyte proliferation in vitro and in vivo using CFSE-like Immunity. 24:53–64. http://dx.doi.org/10.1016/j.immuni.2005.11.012 fluorescent dyes. J. Immunol. Methods. 379:1–14. http://dx.doi.org/ Terry, N.H., and R.A. White. 2006. Flow cytometry after bromodeoxy- 10.1016/j.jim.2012.02.012 uridine labeling to measure S and G2+M phase durations plus dou- Reichard, P. 1988. Interactions between deoxyribonucleotide and DNA bling times in vitro and in vivo. Nat. Protoc. 1:859–869. http://dx.doi synthesis. Annu. Rev. Biochem. 57:349–374. http://dx.doi.org/10.1146/ .org/10.1038/nprot.2006.113 annurev.bi.57.070188.002025 Toy, G., W.R. Austin, H.I. Liao, D. Cheng, A. Singh, D.O. Campbell, Reichard, P. 1993. From RNA to DNA, why so many ribonucleotide re- T.O. Ishikawa, L.W. Lehmann, N. Satyamurthy, M.E. Phelps, et al. ductases? Science. 260:1773–1777. http://dx.doi.org/10.1126/science 2010. Requirement for deoxycytidine kinase in T and B lymphocyte .8511586 development. Proc. Natl. Acad. Sci. USA. 107:5551–5556. http://dx.doi Sabini, E., S. Hazra, S. Ort, M. Konrad, and A. Lavie. 2008. Structural basis .org/10.1073/pnas.0913900107 for substrate promiscuity of dCK. J. Mol. Biol. 378:607–621. http:// Voet, D., and J.G. Voet. 2004. Biochemistry. J. Wiley & Sons, Hoboken, dx.doi.org/10.1016/j.jmb.2008.02.061 NJ. Staub, M., T. Spasokukotskaja, M. Benczur, and F. Antoni. 1988. DNA Xeros, N. 1962. Deoxyriboside control and synchronization of mitosis. synthesis and nucleoside metabolism in human tonsillar lymphocyte Nature. 194:682–683. http://dx.doi.org/10.1038/194682a0 subpopulations. Acta Otolaryngol. Suppl. 454:118–124. http://dx.doi Zhou, X., N. Solaroli, M. Bjerke, J.B. Stewart, B. Rozell, M. Johansson, .org/10.3109/00016488809125014 and A. Karlsson. 2008. Progressive loss of mitochondrial DNA in Taghon, T.N., E.S. David, J.C. Zúñiga-Pflücker, and E.V. Rothenberg. thymidine kinase 2-deficient mice. Hum. Mol. Genet. 17:2329–2335. 2005. Delayed, asynchronous, and reversible T-lineage specification in- http://dx.doi.org/10.1093/hmg/ddn133 duced by Notch/Delta signaling. Genes Dev. 19:965–978. http://dx.doi Zhou, X., M. Johansson, N. Solaroli, B. Rozell, A. Grandien, and A. .org/10.1101/gad.1298305 Karlsson. 2010. Hematopoiesis in the thymidine kinase 2 deficient mouse Taghon, T., M.A. Yui, R. Pant, R.A. Diamond, and E.V. Rothenberg. model of mitochondrial DNA depletion syndrome. J. Inherit. Metab. 2006. Developmental and molecular characterization of emerging Dis. 33:231–236. http://dx.doi.org/10.1007/s10545-010-9102-x

2228 Regulation of replication stress by dCK and TK1 | Austin et al.